[lldb-server][linux] Add ability to allocate memory

This patch adds support for the _M and _m gdb-remote packets, which
(de)allocate memory in the inferior. This works by "injecting" a
m(un)map syscall into the inferior. This consists of:
- finding an executable page of memory
- writing the syscall opcode to it
- setting up registers according to the os syscall convention
- single stepping over the syscall

The advantage of this approach over calling the mmap function is that
this works even in case the mmap function is buggy or unavailable. The
disadvantage is it is more platform-dependent, which is why this patch
only works on X86 (_32 and _64) right now. Adding support for other
linux architectures should be easy and consist of defining the
appropriate syscall constants. Adding support for other OSes depends on
the its ability to do a similar trick.

Differential Revision: https://reviews.llvm.org/D89124

Author: labath

lldb/include/lldb/Host/common/NativeProcessProtocol.h

@@ -17,6 +17,7 @@
 #include "lldb/Utility/ArchSpec.h"
 #include "lldb/Utility/Status.h"
 #include "lldb/Utility/TraceOptions.h"
+#include "lldb/Utility/UnimplementedError.h"
 #include "lldb/lldb-private-forward.h"
 #include "lldb/lldb-types.h"
 #include "llvm/ADT/ArrayRef.h"
@@ -112,10 +113,14 @@ class NativeProcessProtocol {
   virtual Status WriteMemory(lldb::addr_t addr, const void *buf, size_t size,
                              size_t &bytes_written) = 0;
 
-  virtual Status AllocateMemory(size_t size, uint32_t permissions,
-                                lldb::addr_t &addr) = 0;
+  virtual llvm::Expected<lldb::addr_t> AllocateMemory(size_t size,
+                                                      uint32_t permissions) {
+    return llvm::make_error<UnimplementedError>();
+  }
 
-  virtual Status DeallocateMemory(lldb::addr_t addr) = 0;
+  virtual llvm::Error DeallocateMemory(lldb::addr_t addr) {
+    return llvm::make_error<UnimplementedError>();
+  }
 
   virtual lldb::addr_t GetSharedLibraryInfoAddress() = 0;
 

lldb/source/Plugins/Process/FreeBSDRemote/NativeProcessFreeBSD.cpp

@@ -543,15 +543,6 @@ Status NativeProcessFreeBSD::PopulateMemoryRegionCache() {
   return Status();
 }
 
-Status NativeProcessFreeBSD::AllocateMemory(size_t size, uint32_t permissions,
-                                            lldb::addr_t &addr) {
-  return Status("Unimplemented");
-}
-
-Status NativeProcessFreeBSD::DeallocateMemory(lldb::addr_t addr) {
-  return Status("Unimplemented");
-}
-
 lldb::addr_t NativeProcessFreeBSD::GetSharedLibraryInfoAddress() {
   // punt on this for now
   return LLDB_INVALID_ADDRESS;

lldb/source/Plugins/Process/FreeBSDRemote/NativeProcessFreeBSD.h

@@ -60,11 +60,6 @@ class NativeProcessFreeBSD : public NativeProcessELF {
   Status WriteMemory(lldb::addr_t addr, const void *buf, size_t size,
                      size_t &bytes_written) override;
 
-  Status AllocateMemory(size_t size, uint32_t permissions,
-                        lldb::addr_t &addr) override;
-
-  Status DeallocateMemory(lldb::addr_t addr) override;
-
   lldb::addr_t GetSharedLibraryInfoAddress() override;
 
   size_t UpdateThreads() override;

lldb/source/Plugins/Process/Linux/NativeProcessLinux.cpp

@@ -19,6 +19,10 @@
 #include <string>
 #include <unordered_map>
 
+#include "NativeThreadLinux.h"
+#include "Plugins/Process/POSIX/ProcessPOSIXLog.h"
+#include "Plugins/Process/Utility/LinuxProcMaps.h"
+#include "Procfs.h"
 #include "lldb/Core/EmulateInstruction.h"
 #include "lldb/Core/ModuleSpec.h"
 #include "lldb/Host/Host.h"
@@ -38,15 +42,11 @@
 #include "lldb/Utility/State.h"
 #include "lldb/Utility/Status.h"
 #include "lldb/Utility/StringExtractor.h"
+#include "llvm/ADT/ScopeExit.h"
 #include "llvm/Support/Errno.h"
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Threading.h"
 
-#include "NativeThreadLinux.h"
-#include "Plugins/Process/POSIX/ProcessPOSIXLog.h"
-#include "Plugins/Process/Utility/LinuxProcMaps.h"
-#include "Procfs.h"
-
 #include <linux/unistd.h>
 #include <sys/socket.h>
 #include <sys/syscall.h>
@@ -1347,43 +1347,134 @@ void NativeProcessLinux::DoStopIDBumped(uint32_t newBumpId) {
   m_mem_region_cache.clear();
 }
 
-Status NativeProcessLinux::AllocateMemory(size_t size, uint32_t permissions,
-                                          lldb::addr_t &addr) {
-// FIXME implementing this requires the equivalent of
-// InferiorCallPOSIX::InferiorCallMmap, which depends on functional ThreadPlans
-// working with Native*Protocol.
-#if 1
-  return Status("not implemented yet");
-#else
-  addr = LLDB_INVALID_ADDRESS;
-
-  unsigned prot = 0;
-  if (permissions & lldb::ePermissionsReadable)
-    prot |= eMmapProtRead;
-  if (permissions & lldb::ePermissionsWritable)
-    prot |= eMmapProtWrite;
-  if (permissions & lldb::ePermissionsExecutable)
-    prot |= eMmapProtExec;
-
-  // TODO implement this directly in NativeProcessLinux
-  // (and lift to NativeProcessPOSIX if/when that class is refactored out).
-  if (InferiorCallMmap(this, addr, 0, size, prot,
-                       eMmapFlagsAnon | eMmapFlagsPrivate, -1, 0)) {
-    m_addr_to_mmap_size[addr] = size;
-    return Status();
-  } else {
-    addr = LLDB_INVALID_ADDRESS;
-    return Status("unable to allocate %" PRIu64
-                  " bytes of memory with permissions %s",
-                  size, GetPermissionsAsCString(permissions));
+llvm::Expected<uint64_t>
+NativeProcessLinux::Syscall(llvm::ArrayRef<uint64_t> args) {
+  PopulateMemoryRegionCache();
+  auto region_it = llvm::find_if(m_mem_region_cache, [](const auto &pair) {
+    return pair.first.GetExecutable() == MemoryRegionInfo::eYes;
+  });
+  if (region_it == m_mem_region_cache.end())
+    return llvm::createStringError(llvm::inconvertibleErrorCode(),
+                                   "No executable memory region found!");
+
+  addr_t exe_addr = region_it->first.GetRange().GetRangeBase();
+
+  NativeThreadLinux &thread = *GetThreadByID(GetID());
+  assert(thread.GetState() == eStateStopped);
+  NativeRegisterContextLinux &reg_ctx = thread.GetRegisterContext();
+
+  NativeRegisterContextLinux::SyscallData syscall_data =
+      *reg_ctx.GetSyscallData();
+
+  DataBufferSP registers_sp;
+  if (llvm::Error Err = reg_ctx.ReadAllRegisterValues(registers_sp).ToError())
+    return std::move(Err);
+  auto restore_regs = llvm::make_scope_exit(
+      [&] { reg_ctx.WriteAllRegisterValues(registers_sp); });
+
+  llvm::SmallVector<uint8_t, 8> memory(syscall_data.Insn.size());
+  size_t bytes_read;
+  if (llvm::Error Err =
+          ReadMemory(exe_addr, memory.data(), memory.size(), bytes_read)
+              .ToError()) {
+    return std::move(Err);
   }
-#endif
+
+  auto restore_mem = llvm::make_scope_exit(
+      [&] { WriteMemory(exe_addr, memory.data(), memory.size(), bytes_read); });
+
+  if (llvm::Error Err = reg_ctx.SetPC(exe_addr).ToError())
+    return std::move(Err);
+
+  for (const auto &zip : llvm::zip_first(args, syscall_data.Args)) {
+    if (llvm::Error Err =
+            reg_ctx
+                .WriteRegisterFromUnsigned(std::get<1>(zip), std::get<0>(zip))
+                .ToError()) {
+      return std::move(Err);
+    }
+  }
+  if (llvm::Error Err = WriteMemory(exe_addr, syscall_data.Insn.data(),
+                                    syscall_data.Insn.size(), bytes_read)
+                            .ToError())
+    return std::move(Err);
+
+  m_mem_region_cache.clear();
+
+  // With software single stepping the syscall insn buffer must also include a
+  // trap instruction to stop the process.
+  int req = SupportHardwareSingleStepping() ? PTRACE_SINGLESTEP : PTRACE_CONT;
+  if (llvm::Error Err =
+          PtraceWrapper(req, thread.GetID(), nullptr, nullptr).ToError())
+    return std::move(Err);
+
+  int status;
+  ::pid_t wait_pid = llvm::sys::RetryAfterSignal(-1, ::waitpid, thread.GetID(),
+                                                 &status, __WALL);
+  if (wait_pid == -1) {
+    return llvm::errorCodeToError(
+        std::error_code(errno, std::generic_category()));
+  }
+  assert((unsigned)wait_pid == thread.GetID());
+
+  uint64_t result = reg_ctx.ReadRegisterAsUnsigned(syscall_data.Result, -ESRCH);
+
+  // Values larger than this are actually negative errno numbers.
+  uint64_t errno_threshold =
+      (uint64_t(-1) >> (64 - 8 * m_arch.GetAddressByteSize())) - 0x1000;
+  if (result > errno_threshold) {
+    return llvm::errorCodeToError(
+        std::error_code(-result & 0xfff, std::generic_category()));
+  }
+
+  return result;
 }
 
-Status NativeProcessLinux::DeallocateMemory(lldb::addr_t addr) {
-  // FIXME see comments in AllocateMemory - required lower-level
-  // bits not in place yet (ThreadPlans)
-  return Status("not implemented");
+llvm::Expected<addr_t>
+NativeProcessLinux::AllocateMemory(size_t size, uint32_t permissions) {
+
+  llvm::Optional<NativeRegisterContextLinux::MmapData> mmap_data =
+      GetCurrentThread()->GetRegisterContext().GetMmapData();
+  if (!mmap_data)
+    return llvm::make_error<UnimplementedError>();
+
+  unsigned prot = PROT_NONE;
+  assert((permissions & (ePermissionsReadable | ePermissionsWritable |
+                         ePermissionsExecutable)) == permissions &&
+         "Unknown permission!");
+  if (permissions & ePermissionsReadable)
+    prot |= PROT_READ;
+  if (permissions & ePermissionsWritable)
+    prot |= PROT_WRITE;
+  if (permissions & ePermissionsExecutable)
+    prot |= PROT_EXEC;
+
+  llvm::Expected<uint64_t> Result =
+      Syscall({mmap_data->SysMmap, 0, size, prot, MAP_ANONYMOUS | MAP_PRIVATE,
+               uint64_t(-1), 0});
+  if (Result)
+    m_allocated_memory.try_emplace(*Result, size);
+  return Result;
+}
+
+llvm::Error NativeProcessLinux::DeallocateMemory(lldb::addr_t addr) {
+  llvm::Optional<NativeRegisterContextLinux::MmapData> mmap_data =
+      GetCurrentThread()->GetRegisterContext().GetMmapData();
+  if (!mmap_data)
+    return llvm::make_error<UnimplementedError>();
+
+  auto it = m_allocated_memory.find(addr);
+  if (it == m_allocated_memory.end())
+    return llvm::createStringError(llvm::errc::invalid_argument,
+                                   "Memory not allocated by the debugger.");
+
+  llvm::Expected<uint64_t> Result =
+      Syscall({mmap_data->SysMunmap, addr, it->second});
+  if (!Result)
+    return Result.takeError();
+
+  m_allocated_memory.erase(it);
+  return llvm::Error::success();
 }
 
 size_t NativeProcessLinux::UpdateThreads() {
@@ -1652,6 +1743,11 @@ NativeThreadLinux *NativeProcessLinux::GetThreadByID(lldb::tid_t tid) {
       NativeProcessProtocol::GetThreadByID(tid));
 }
 
+NativeThreadLinux *NativeProcessLinux::GetCurrentThread() {
+  return static_cast<NativeThreadLinux *>(
+      NativeProcessProtocol::GetCurrentThread());
+}
+
 Status NativeProcessLinux::ResumeThread(NativeThreadLinux &thread,
                                         lldb::StateType state, int signo) {
   Log *const log = ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_THREAD);

lldb/source/Plugins/Process/Linux/NativeProcessLinux.h

@@ -71,10 +71,10 @@ class NativeProcessLinux : public NativeProcessELF {
   Status WriteMemory(lldb::addr_t addr, const void *buf, size_t size,
                      size_t &bytes_written) override;
 
-  Status AllocateMemory(size_t size, uint32_t permissions,
-                        lldb::addr_t &addr) override;
+  llvm::Expected<lldb::addr_t> AllocateMemory(size_t size,
+                                              uint32_t permissions) override;
 
-  Status DeallocateMemory(lldb::addr_t addr) override;
+  llvm::Error DeallocateMemory(lldb::addr_t addr) override;
 
   size_t UpdateThreads() override;
 
@@ -94,6 +94,7 @@ class NativeProcessLinux : public NativeProcessELF {
                             lldb::addr_t &load_addr) override;
 
   NativeThreadLinux *GetThreadByID(lldb::tid_t id);
+  NativeThreadLinux *GetCurrentThread();
 
   llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>>
   GetAuxvData() const override {
@@ -127,6 +128,8 @@ class NativeProcessLinux : public NativeProcessELF {
   llvm::Expected<llvm::ArrayRef<uint8_t>>
   GetSoftwareBreakpointTrapOpcode(size_t size_hint) override;
 
+  llvm::Expected<uint64_t> Syscall(llvm::ArrayRef<uint64_t> args);
+
 private:
   MainLoop::SignalHandleUP m_sigchld_handle;
   ArchSpec m_arch;
@@ -140,6 +143,9 @@ class NativeProcessLinux : public NativeProcessELF {
   // the relevan breakpoint
   std::map<lldb::tid_t, lldb::addr_t> m_threads_stepping_with_breakpoint;
 
+  /// Inferior memory (allocated by us) and its size.
+  llvm::DenseMap<lldb::addr_t, lldb::addr_t> m_allocated_memory;
+
   // Private Instance Methods
   NativeProcessLinux(::pid_t pid, int terminal_fd, NativeDelegate &delegate,
                      const ArchSpec &arch, MainLoop &mainloop,

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux.h

@@ -31,6 +31,32 @@ class NativeRegisterContextLinux : public NativeRegisterContextRegisterInfo {
   // Invalidates cached values in register context data structures
   virtual void InvalidateAllRegisters(){}
 
+  struct SyscallData {
+    /// The syscall instruction. If the architecture uses software
+    /// single-stepping, the instruction should also be followed by a trap to
+    /// ensure the process is stopped after the syscall.
+    llvm::ArrayRef<uint8_t> Insn;
+
+    /// Registers used for syscall arguments. The first register is used to
+    /// store the syscall number.
+    llvm::ArrayRef<uint32_t> Args;
+
+    uint32_t Result; ///< Register containing the syscall result.
+  };
+  /// Return architecture-specific data needed to make inferior syscalls, if
+  /// they are supported.
+  virtual llvm::Optional<SyscallData> GetSyscallData() { return llvm::None; }
+
+  struct MmapData {
+    // Syscall numbers can be found (e.g.) in /usr/include/asm/unistd.h for the
+    // relevant architecture.
+    unsigned SysMmap;   ///< mmap syscall number.
+    unsigned SysMunmap; ///< munmap syscall number
+  };
+  /// Return the architecture-specific data needed to make mmap syscalls, if
+  /// they are supported.
+  virtual llvm::Optional<MmapData> GetMmapData() { return llvm::None; }
+
 protected:
   lldb::ByteOrder GetByteOrder() const;
 

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux_x86_64.cpp

@@ -1220,4 +1220,38 @@ NativeRegisterContextLinux_x86_64::GetPtraceOffset(uint32_t reg_index) {
          (IsMPX(reg_index) ? 128 : 0);
 }
 
+llvm::Optional<NativeRegisterContextLinux::SyscallData>
+NativeRegisterContextLinux_x86_64::GetSyscallData() {
+  switch (GetRegisterInfoInterface().GetTargetArchitecture().GetMachine()) {
+  case llvm::Triple::x86: {
+    static const uint8_t Int80[] = {0xcd, 0x80};
+    static const uint32_t Args[] = {lldb_eax_i386, lldb_ebx_i386, lldb_ecx_i386,
+                                    lldb_edx_i386, lldb_esi_i386, lldb_edi_i386,
+                                    lldb_ebp_i386};
+    return SyscallData{Int80, Args, lldb_eax_i386};
+  }
+  case llvm::Triple::x86_64: {
+    static const uint8_t Syscall[] = {0x0f, 0x05};
+    static const uint32_t Args[] = {
+        lldb_rax_x86_64, lldb_rdi_x86_64, lldb_rsi_x86_64, lldb_rdx_x86_64,
+        lldb_r10_x86_64, lldb_r8_x86_64,  lldb_r9_x86_64};
+    return SyscallData{Syscall, Args, lldb_rax_x86_64};
+  }
+  default:
+    llvm_unreachable("Unhandled architecture!");
+  }
+}
+
+llvm::Optional<NativeRegisterContextLinux::MmapData>
+NativeRegisterContextLinux_x86_64::GetMmapData() {
+  switch (GetRegisterInfoInterface().GetTargetArchitecture().GetMachine()) {
+  case llvm::Triple::x86:
+    return MmapData{192, 91};
+  case llvm::Triple::x86_64:
+    return MmapData{9, 11};
+  default:
+    llvm_unreachable("Unhandled architecture!");
+  }
+}
+
 #endif // defined(__i386__) || defined(__x86_64__)

lldb/source/Plugins/Process/Linux/NativeRegisterContextLinux_x86_64.h

@@ -64,6 +64,10 @@ class NativeRegisterContextLinux_x86_64 : public NativeRegisterContextLinux {
 
   uint32_t NumSupportedHardwareWatchpoints() override;
 
+  llvm::Optional<SyscallData> GetSyscallData() override;
+
+  llvm::Optional<MmapData> GetMmapData() override;
+
 protected:
   void *GetGPRBuffer() override { return &m_gpr_x86_64; }